The diet of pinnipeds is most commonly inferred from morphologically diagnostic remains of prey in their scats. Although this method can generate quantitative estimates of diet simply, important prey types may not always be detected. DNA-based methods improve detection of prey in scats, but they are not quantitative. While some studies have combined morphological and DNA-based methods, these have only assessed prey that are represented by their hard remains in scats. To overcome this bias, we apply molecular and morphological analyses to the soft and hard portions of faecal samples respectively, to estimate the diet of lactating Antarctic fur seals (Arctocephalus gazella) on Heard Island. The diet of this population is of particular interest because it is expanding rapidly and may rely to some extent on mackerel icefish (Champsocephalus gunnari), which are subject to commercial fisheries. Based on results from morphological analysis and likely important prey types, we tested for DNA remains of C. gunnari, myctophids and squid in faecal samples. The proportion of samples (n = 54) yielding no dietary information was reduced from around 25.9% using either method alone, to 9.3% when combined. Detection of all prey types tested for was notably improved by integrating molecular and morphological data. Data from either method alone would have underestimated the number of animals consuming C. gunnari by around 25.7%. Detection of multiple prey types in samples increased from 9.3% when using morphological analysis only, to 33.3% when using DNA only, to 46.3% when using both methods. Taken in isolation, morphological data inferred that individual seals consume either C. gunnari or myctophids, probably foraging in separate locations characteristic of those prey. Including molecular data demonstrated that while this may be true of some individuals, many other seals consume a mixed diet of at least C. gunnari, myctophids and squid. This new approach of combining DNA-based and morphological analyses of diet samples markedly increased the number of samples yielding dietary information, as well as increasing the amount of information attained from those samples. Our findings illustrate the broad potential of this technique to improve insight into trophic interactions in marine ecosystems.